Storm Prediction Center Highlights
NCEP Production Suite Review
December 2, 2014
Steven Weiss, Israel Jirak, Chris Melick, Andy Dean, Patrick Marsh, and
Russ Schneider
Storm Prediction Center, Norman, OK
National Weather Center
Outline
• Good news and thanks from SPC
• SPC mission and responsibility
• Convection-allowing model (CAM) forecasts
for some high-impact 2014 severe weather
cases
– Focus on NAM Nest and HiResW upgrades and
HRRR implementation
• CAM ensemble systems
– How do we move forward operationally as
experimental CAM ensembles continue to
proliferate?
Thanks From SPC Perspective
• Continued very reliable, timely operational model production
• Forums for enhanced collaboration are paying dividends
• Weekly EMC Model Evaluation Group (MEG) webinars
• Bi-Weekly/Monthly EMC-SPC telecons/webinars focusing on convectionallowing model improvements (NAM Nest, HiResWindow)
• Bi-Weekly GSD-SPC telecons/webinars focusing on RAP and HRRR
development
• Improvements in Convection-Allowing Models for severe weather
forecasting support
• NAM Nest, HiResW NMMB and ARW upgrades
• Operational HRRR implementation
• Closer integration of model development efforts with severe weather
forecasting needs have been occurring – need to maintain momentum
• Consistent with NOAA/NWS strategic goals of Weather Ready Nation,
Warn on Forecast, and emerging FACETs (Forecasting A Continuum of
Environmental Threats (FACETs)
STORM PREDICTION CENTER
HAZARDOUS PHENOMENA
• Thunderstorms
• Hail, Wind, Tornadoes
• Fire Weather
•
Mesoscale Winter Weather
StormSPC
Prediction
Primary Products
MissionCenter
and Responsibility
Convection-Allowing Model (CAM)
Guidance at SPC
• SPC forecasters have had access to CAM guidance since 2004
– Hazardous Weather Testbed Spring Forecasting Experiment evaluations
– Year-round use in SPC operational severe weather forecasting
– Primary partnerships with NCEP/EMC, NSSL, ESRL/GSD, OU/CAPS, NCAR,
and AFWA
• Typical CAM model characteristics
–
–
–
–
Horizontal grid length ~3-4 km
No parameterized convection
Often “cold-start” from NAM or larger-scale model initial conditions
Since 2007 pioneering work on real-time CAM ensemble systems
• CAMs provide unique convective storm forecast guidance
– Simulated reflectivity & synthetic satellite imagery
– Extraction of storm characteristics using Hourly Max Fields (HMFs)
• Updraft helicity, updraft/downdraft speed, 10-m wind, vertically integrated graupel
Flashback – NPSR December 2012
Super Outbreak 27-28 April 2011
Images Valid 00z 28 April
NAM 4 km Nest
24-hr forecast
of 1 km AGL
Simulated
Reflectivity
Observed
Radar Base
Reflectivity
Mosaic
Very few intense discrete storms
predicted; model storms are too
broad/weak and lack realism
Observed radar displayed numerous
discrete supercells and several
embedded bow echoes
Action: Started SPC-EMC bi-weekly telecons/webinars in Jan. 2013
Goal: Improve convective storm structure, mode and intensity
Summary of Recent EMC
Convection-Allowing Model Upgrades
Characteristics
HiResW NMMB
HiResW ARW
NAM Nest
Implementation
Date
11 June 2014
11 June 2014
12 August 2014
Horizontal Grid
3.6 km
4.2 km
4.0 km
Vertical Levels
40
40
60
Initial Conditions
RAP
RAP
NAM
Microphysics
Updated Ferrier
WSM6
Ferrier-Aligo
PBL/Turbulence
MYJ
MYJintensity so
A key goal: Improve
convective storm YSU
structure, mode and
simulated model storms are more similar to those observed by radar
Radiation
RRTM
RRTM
RRTM
Conv. Param.
None
None
None
Domain
CONUS
CONUS
CONUS
Forecast Length
48 hrs
48 hrs
60 hrs
Run Times
00, 12z
00, 12z
00, 06,12,18z
Simulated Composite Reflectivity
22-hr Forecasts Valid 22z 20 May 2013
Vertical Cross-Section (top); Plan View (bottom)
Experiments with New Ferrier-Aligo Microphysics => More intense, realistic storms
April 27, 2014
• High Risk Over Parts of AR
– Significant Tornadoes in AR and MS including one EF-4 tornado
– 19 Fatalities including 16 in Mayflower-Vilonia AR (41 mile path length)
• Deadliest Single Tornado in AR Since May 15, 1968
Destroyed Homes in Mayflower, AR
NAM Nest Comparison
24-hr Forecast Valid 00z 28 April 2014
NAM Nest
Parallel
Observed
Radar
NAM Nest
Operational
• Parallel NAM Nest shows more
intense, smaller-scale storm details
• Most intense storms correctly
predicted over central AR by parallel,
whereas operational run has broader
region of strong storms
HiResW NMMB Comparison
24-hr Forecast Valid 00z 28 April 2014
HiResW
NMMB
Parallel
Observed
Radar
HiResW
WRF-NMM
Operational
• Neither HiResW NMMB nor WRFNMM highlighted central AR region
• Both have limited storm coverage
• NMMB parallel focuses too far west
• WRF-NMM operational too far east
HiResW ARW Comparison
24-hr Forecast Valid 00z 28 April 2014
HiResW
ARW
Parallel
Observed
Radar
HiResW
ARW
Operational
• Parallel HiResW ARW has more
intense storms, especially over
eastern AR (but too far east)
• Parallel considered an improvement
over operational, which focused
weaker storms over MO
April 28, 2014
• High Risk Over Parts of MS and AL
– Significant Tornadoes in MS, AL and TN including EF-4 with 34 mile
path length
– 15 Fatalities (10 in MS)
Also Hail to Baseball Size
Tornado Near Louisville, MS
NAM Nest Comparison
22-hr Forecast Valid 22z 28 April 2014
NAM Nest
Parallel
Observed
Radar
NAM Nest
Operational
• Parallel NAM Nest shows more
intense storms and is more similar to
observed radar
• Both runs predict NE-SW oriented
band of storms
HiResW NMMB Comparison
22-hr Forecast Valid 22z 28 April 2014
HiResW
NMMB
Parallel
Observed
Radar
HiResW
WRF-NMM
Operational
• Neither run predicts primary linear
mode
• Parallel HiResW NMMB has
improved storm details and is better
indicating threat over northern MS,
but is too slow with eastward
progression
HiResW ARW Comparison
22-hr Forecast Valid 22z 28 April 2014
HiResW
ARW
Parallel
Observed
Radar
HiResW
ARW
Operational
• Both runs indicate linear structure
displaced to the southeast
• HiResW ARW parallel appears to
develop too strong a cold pool
evidenced by bowing structure
• Storms are more intense in parallel
May 8, 2014
• Moderate Risk Parts of MN and IA
– Wind Damage, Significant Hail, and Weak Tornadoes
– Case Occurred During HWT Spring Forecasting Experiment
House Damage in Red Wing, MN
NAM Nest Comparison
20-hr Forecast Valid 20z 8 May 2014
NAM Nest
Parallel
Observed
Radar
NAM Nest
Operational
• Parallel NAM Nest shows more
intense, smaller-scale storm details
• Improved over Iowa with weaker
reflectivity
• Both runs predict spurious storms
over eastern SD and northern WI
HiResW NMMB Comparison
20-hr Forecast Valid 20z 8 May 2014
HiResW
NMMB
Parallel
Observed
Radar
HiResW
WRF-NMM
Operational
• Parallel HiResW ARW was much
better over MN than operational run
• RAP initial conditions may have
contributed to improved performance
HiResW ARW Comparison
20-hr Forecast Valid 20z 8 May 2014
HiResW
ARW
Parallel
Observed
Radar
HiResW
ARW
Operational
• Parallel HiResW ARW was much
better over MN than operational run
• RAP initial conditions may have
contributed to improved performance
June 3, 2014
• Major Severe Weather Event Over Central Plains and Lower
MO Valley
– Series of Severe Storms Caused Tornadoes and Significant Hail/Wind
– Substantial Flooding parts of NE and IA including Omaha Metro
Hail to Softball Size and Winds to 90 mph
Baseball Hail in Norfolk, NE
NAM Nest Comparison
24-hr Forecast Valid 00z 4 June 2014
NAM Nest
Parallel
Observed
Radar
NAM Nest
Operational
• Parallel NAM Nest shows more
intense, smaller-scale storm details,
but is worse with storm coverage and
NW-SE axis of observed severe storms
HiResW NMMB Comparison
24-hr Forecast Valid 00z 4 June 2014
HiResW
NMMB
Parallel
Observed
Radar
HiResW
WRF-NMM
Operational
• Parallel HiResW NMMB shows
smaller-scale storms, but similar
location and coverage
• Neither run captures the NW-SE axis
of severe storms
HiResW ARW Comparison
24-hr Forecast Valid 00z 4 June 2014
HiResW
ARW
Parallel
Observed
Radar
HiResW
ARW
Operational
• Parallel HiResW ARW is much
improved with intensity, location, and
coverage of storms within the severe
storm axis
June 6, 2014
• Severe Weather Event Over Central & Southern High Plains
– Bowing Squall Lines Moved Eastward Producing Tornadoes and
Significant Hail & Wind in CO, NM, NE, KS, and TX
– Last Day of HWT Spring Forecasting Experiment
Hail to Baseball Size and Winds to 85 mph
Storm near Scott City, KS
NAM Nest Comparison
27-hr Forecast Valid 03z 7 June 2014
NAM Nest
Parallel
Observed
Radar
NAM Nest
Operational
• Parallel NAM Nest shows smallerscale storm details
• Parallel and operational runs show
similar placement and coverage of
predicted storms, with north-south
extent too limited
HiResW NMMB Comparison
27-hr Forecast Valid 03z 7 June 2014
HiResW
NMMB
Parallel
Observed
Radar
HiResW
WRF-NMM
Operational
• Parallel HiResW NMMB has more
limited storm coverage
• Operational run provided better
overall guidance despite
displacement error to the west
HiResW ARW Comparison
27-hr Forecast Valid 03z 7 June 2014
HiResW
ARW
Parallel
Observed
Radar
HiResW
ARW
Operational
• Similarly, parallel HiResW ARW
storm coverage was more limited in
north-south extent
• Operational run provided better
overall guidance
Subjective Comparison of
Operational and Parallel Runs During
HWT Spring Forecasting Experiment
NAM Nest
HiResW ARW
Parallel Better
Parallel Better
HiResW NMMB
Focusing on reflectivity and UH
forecasts for severe weather
guidance, parallel runs subjectively
rated as good as or better than the
operational model:
HiResW ARW – 18 out of 23 days
HiResW NMMB – 21 out of 23 days
NAM Nest – 17 out of 23 days
June 16, 2014
• Significant Severe Weather Event Over Midwest
• At least four EF-4 Tornadoes in Northeast Nebraska
– Only 2 Fatalities in Sparsely Populated Area Despite Direct Hit of
Pilger, NE
Hail to Softball Size and Winds to 95 mph
Twin Tornadoes in Pilger, NE
NAM Nest and HiResW
21-hr Forecast Valid 21z 16 June 2014
NAM Nest
HiResW
NMMB
Observed
Radar
HiResW
ARW
Pilger
Storm
NAM Nest and HiResW
24-hr Forecast Valid 00z 17 June 2014
NAM Nest
HiResW
NMMB
• NAM Nest Performed Well with Primary Severe Storm Complex But Missed Pilger, NE Storm
• HiResW Runs Performed Poorly and Did Not Predict Observed Convective Storm Evolution
Observed
Radar
HiResW
ARW
June 30, 2014
• Significant Severe Weather Event Over Midwest
• Two Derecho-Producing MCSs Tracked Along Similar Paths
– Systems Moved at 50-60 kt Forward Speed
– Widespread Wind Damage and Several EF-1 Tornadoes
Hail to Baseball Size and Winds to 90 mph
Lightning Time Lapse Over Chicago
June 30, 2014
System 1
Chicago was hit by two derecho
systems over a 4-hour period.
The first struck around 00z and
the second between 03-04z.
We’ll look at the 24- and 27-hr
forecasts valid at 00z and 03z.
System 2
NAM Nest and HiResW
24-hr Forecast Valid 00z 1 July 2014
NAM Nest
HiResW
NMMB
• Models did not capture well Derecho 1 although NAM Nest did show a QLCS across WI
• NAM Nest correctly indicated a second QLCS across eastern IA; HiResW runs were slow
developing 2nd system
Observed
Radar
HiResW
ARW
Derecho 1
Derecho 2
NAM Nest and HiResW
27-hr Forecast Valid 03z 1 July 2014
NAM Nest
HiResW
NMMB
• NAM Nest was quite good predicting second QLCS
• HiResW runs “caught up” to some extent but were slow with northern part of system
Observed
Radar
HiResW
ARW
HiResW & NAM Nest Objective Verification
Performance Diagram – Reflectivity
•
Neighborhood verification
(ROI=40 km) of reflectivity
(13 Aug 2014 – 22 Nov 2014)
o Period after new NAM
implementation
o Objective verification of
reflectivity forecasts using a
spatial neighborhood method
o Performance of NAM Nest and
HiResW runs is similar
• HiResW ARW has slightly
higher POD/CSI and lower
FAR for 30 & 40 dBZ
thresholds
• Bias ~1 for all models
o Indicates forecasts are equally
plausible in statistical sense
• Useful for CAM ensemble
construction (SSEO)
Higher
POD
Lower FAR
Summary: NAM Nest & HiResW
• New HiResW and NAM Nest models can provide better
severe storm guidance compared to previous versions
– Improved storm structure, mode, and intensity
– HiResW - Full CONUS domain at 00 and 12z
– HiResW - Use of RAP initial conditions enhances diversity in CAM suite
HiResW - Increased horizontal/vertical resolution
– NAM Nest and HiResW – Updated microphysics, diffusion, radiation
• But, considerable challenges remain in convective-scale
prediction for high-impact weather
– Convective storm prediction is very hard to do well (where, when, etc.)
– Foundational for Warn on Forecast, FACETs & Weather Ready Nation
– Challenges include: Observational data, DA, model dynamic core and
physics, information extraction, post-processing, and display, etc.
Use of HRRR Model Guidance at SPC
• SPC forecasters have had access to ESRL HRRR grids since
April 2010
– Also evaluated in HWT Spring Forecasting Experiments
– No surprise – HRRR is useful for short-term forecast applications
• The NCEP version of the HRRR has a higher reflectivity bias
than the ESRL version, but the timeliness and reliability is
much improved for the NCEP HRRR
• Concept of HRRR run-to-run continuity leading to higher
forecaster confidence must be used with caution
̶
Implied spread-skill relationship is better in strongly forced
situations
• Issues with mixing in the warm-sector PBL lead to conditions
that are too warm/dry and occasionally spurious convective
initiation
HRRR – September 20, 2014
Strongly Forced Case in Upper Midwest
Consistent Model Forecasts Over Many Cycles
Low-Topped Band of Storms in Cool, Low Dew Point Environment
Mainly Wind Damage in Minnesota
Storm Clouds Near Crookston, MN
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
12z Cycle: 9-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
13z Cycle: 8-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
14z Cycle: 7-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
15z Cycle: 6-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
16z Cycle: 5-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
17z Cycle: 4-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
18z Cycle: 3-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
Forecast
19z Cycle: 2-hr HRRR Forecast
HRRR Run-to-Run Continuity
Strongly Forced Case (“Comma Shape”)
Images Valid 21z 20 September 2014
Observed Radar
20z Cycle: 1-hr HRRR Forecast
Forecast
In strongly forced situations, can have more confidence in consistent solutions
HRRR Predicts Spurious Evening
Convection over Kansas
Five Consecutive Hourly ESRL HRRR Forecasts
Initialized 18-22z
Valid 23-06z on 17-18 September 2014
HRRR Forecasts valid 23z 17 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
OBSERVED RADAR
HRRR Forecasts valid 00z 18 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
OBSERVED RADAR
HRRR Forecasts valid 01z 18 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
OBSERVED RADAR
HRRR Forecasts valid 02z 18 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
OBSERVED RADAR
HRRR Forecasts valid 03z 18 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
OBSERVED RADAR
HRRR Forecasts valid 04z 18 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
OBSERVED RADAR
HRRR Forecasts valid 05z 18 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
OBSERVED RADAR
HRRR Forecasts valid 06z 18 Sep 2014
Consecutive Model Runs from 18z, 19z, 20z, 21z and 22z
Summary: HRRR runs consistently predict spurious MCS rather than
the elevated convective band to the northeast
OBSERVED RADAR
Summary: HRRR Run-to-Run Consistency
and Forecaster Confidence
• At times, especially during more strong forced situations, runto-run continuity can add to forecaster confidence in the
utility of the HRRR forecasts
• However, there are other times when the HRRR appears to
“lock in” to a solution that turns out to be erroneous
• The “spread/skill” relationship of time-lagged HRRR forecasts
for deep convection can vary substantially
• Forecasters should be judicious when applying traditional
concepts of run-to-run continuity as a measure of confidence
in the forecast accuracy
The New Frontier:
Convection-Allowing Ensembles
• Lower predictability exists at the convective storm scale
– Poor sampling of mesoscale environment (IC uncertainty)
– Model physics errors tend to increase at smaller scales
• A number of experimental convection-allowing ensemble
systems have been developed in recent years
– Information about range of convective storm possibilities
– Probabilistic information on likelihood of convective storm
occurrence, coverage, intensity, and proxy severe storm
attributes
• CAM ensembles are essential to support Weather Ready
Nation high-impact weather needs
2014 Spring Forecasting Experiment
Four Convection-Allowing Ensembles
• OU/CAPS Storm-Scale Ensemble Forecast (SSEF) System
̶
̶
Since 2007; 60-hr forecasts from 00z; 12z runs began in 2013
Primarily WRF-ARW; 4-km grid spacing; forecasts to 60hrs in 2014
Multi-physics, multi-initial conditions: applies SREF perturbations to NAM ICs
Advanced physics, 3DVAR & radar data assimilation; available for HWT/SFE
̶
̶
• SPC Storm-Scale Ensemble of Opportunity (SSEO)
̶
̶
Since 2011; 36-hr forecasts at 00z & 12z; 7 members (2 time-lagged)
Multi-model (ARW, WRF-NMM & NMM-B), multi-physics; ~4-km grid spacing
Uses available deterministic models at SPC to process as an ensemble
Basic data assimilation through NDAS; available year-round in SPC
̶
̶
• Air Force Weather Agency (AFWA) Ensemble
̶
̶
Since 2012; 60-hr forecasts at 00z & 12z; 10 members; 4-km grid spacing
Single model (WRF-ARW), multi-physics, multi-initial conditions
Cold start from downscaled global model forecasts (GFS, UM, CMC)
No data assimilation; available year-round in SPC
̶
̶
• National Severe Storms Laboratory (NSSL) WRF Ensemble
̶
̶
̶
First year; 36-hr forecasts at 00Z; 9 members; 4-km grid spacing
Single model (WRF-ARW), single-physics, multi-initial conditions
Cold start from 3-h forecasts of selected SREF members; year-round in SPC
2014 HWT CAM Ensemble UH Forecasts
Max; Nprob>25;Nprob>100; Valid 21-00z 3 June 2014
CAPS SSEF
NSSL Ens.
SPC SSEO
AFWA Ens.
Max From Any Member
NProb UH>25 m2/s2
NProb UH>100 m2/s2
Despite differences in CAM
ensemble configurations
and degree of complexity
in data assimilation, IC/LBC
perturbations, model and
physics diversity, etc.,
overall performance was
often similar on many days
HWT CAM Ensemble Subjective Evaluation
Hourly Probability of Reflectivity > 40 dBZ by Weighted Average
• Overall scores very similar when applying weighted average across the ratings
• Continues to suggest that CAM ensemble performance is similar regardless of
system configuration and degree of complexity
More Research Efforts on CAM Ensembles
(But wait, there’s more)
• NCAR has been developing an EnKF-based Mesoscale/CAM
Ensemble System since 2011
– Considerably improved by 2013 for MPEX forecasting support
• Goal for 2015: CONUS 10 member, 3 km ARW Ensemble with
forecasts to 48 hrs and run daily in real-time for at least 1 year
– Initialized by downscaling 10 analysis members of a continuously cycling EnKF 15 km
mesoscale system comprised of 50 members
– Blending of ensemble DA and prediction model can result in more dynamically
consistent initialization and forecast process
• NCAR Ensemble expected to part of HWT Spring Experiment
and available for assessment in SPC operations (pending
approval of HPC request)
• This will be CAM ensemble #5 - How many ensembles are
too many?
– Help! Organized community-wide collaboration is clearly needed
More Research Efforts on CAM Ensembles
(But wait, there’s more)
• For CAM ensembles, we are approaching where we were 5-10 years
ago with deterministic CAMs
– To enhance R2O/O2R process, considerable collaboration across the research
and operational communities is needed to identify approaches feasible for
operational implementation of CAM ensemble(s)
– HWT is positioned to play a key role in the assessment process
– Many key issues remain - will require sophisticated verification methods
•
•
•
•
•
•
Multi-model versus single model
Necessary horizontal and vertical resolution
Scale-appropriate sources and perturbation strategies for ICs/LBCs
Physics diversity and level of complexity
Innovative statistical approaches to extract storm-scale information
Creation of reliable probabilistic information for high-impact weather
– Operational computing resources will ultimately inform CAM ensemble
decision process
More Research Efforts on CAM Ensembles
(Provocative Slide to Spur Discussion)
• Effective WCOSS utilization is a key foundational component
• UCACN Recommendation – Simplify operational model suite and
reduce number of modeling systems
– Emulate other international modeling centers that have developed a
more unified modeling system
– Example: As GFS and GEFS approach resolutions of NAM and SREF, are
all four systems needed?
• Consider creating a team composed of operational forecasters, EMC
and research modelers, and stakeholders in government and private
sector to discuss and help plan future modeling suite
– “Think outside the box” but recognize we will always have resource
limits
Two More Slides
HRRR Performance:
Warm-Sector PBL Mixing
• The HRRR (and RAP) have a
tendency to overmix the PBL,
resulting in a boundary layer
that is too warm and dry
relative to observations.
• In this example from 18 June
2014, the HRRR 6-h forecast
of surface T/Td valid at 00z
was 97/63 while observations
showed 89/74.
• The boundary layer becomes
too deep/well-mixed, which
can result in spurious
convective initiation.
18Z NCEP HRRR, F006 Valid 00Z 06-18-14
00Z OBS SOUNDING in purple
HRRR Performance
Issue: Warm-Sector PBL Mixing
6-h forecast Valid 00Z June 18
Radar 00Z June 18
• Overmixing in the warm sector for this case led to an
area of spurious convection across western Iowa and
eastern Nebraska (and it was worse in later runs).
• The RAP/HRRR group at ESRL/GSD are working on
solutions for this issue.